On October 19, 2026, OpenAI announced new progress in improving ChatGPT’s coding abilities for low-resource programming languages. According to the official OpenAI announcement, this work targets languages and tasks that have historically lacked extensive online data and benchmarks. OpenAI specifically mentions improved performance for software engineering tasks, as well as substantive work on two important challenges: evaluating ChatGPT results on the SQLite open-source database, and supporting the automated migration of COBOL legacy systems.
- ChatGPT is now better at answering questions and generating code in languages that are less common in modern repositories.
- The team created new evaluation methodologies to measure generative performance across low-resource language scenarios—including accurate SQL generation for SQLite.
- OpenAI is also pushing toward automated COBOL migration, which can help organizations modernize old IT infrastructure.
If you work in software engineering and deal with legacy systems or maintain codebases in rarely-used languages, these improvements mean you can expect more reliable support from ChatGPT. OpenAI encourages feedback on these evaluation methods to further refine the model’s abilities in real-world applications.